Abstract
Over the last few years more than 400 anomalies have been identified in the literature, motivating the emergence of new factor models. This trend is currently under meticulous re-evaluation. Critics argue that researchers have incentives to publish new factors rather than check the validity of published ones. In this context, we propose a stochastic spanning approach to assess whether a trading strategy constitutes a violation of a given factor model. Compared with the traditional mean-variance spanning approach, our framework is particularly appealing for asset classes and investment strategies with asymmetric risk profiles such as the anomalies examined. Our evidence shows that only few of the examined strategies actually expand the opportunity set of the general risk-averter. Most importantly, we find that stochastic dominance spanning yields significantly more robust results than mean-variance spanning. Our approach contributes to the effort that is undertaken in the literature to re-evaluate published anomalies and discern those with real economic content from those that may be the result of biases and data-snooping.